«

Thanks for the clarification.

I guess it is hard to find some general rule of thumb about which
sensor size can properly handle what level of megapixels.

I guess I was hooping for something like this, for example:

"An Oly 4/3rds sensor is large enough to properly resolve a full
12 megapixels of image data - anything above that and you might
start to lose detail because the sensor cannot properly
accommodate more than 12 megapixels. In order to get higher than
12 megapixels and have those extra pixels be meaningful, they
would have to increase the size of the sensor"

But I guess it isn't that simple.

I do appreciate the reply. The more I read, the less that
article seems to be on target. They generalized too much, I
think.

Canon certainly seems to feel that the sensor in their 10mp A640
is fully capable of handling all 10mp of data properly, and
Olympus seems to feel very confident that the size of its 4/3rd
sensor is able to handle a full 8mp without any compromise in
quality.

Anyway, I'll keep on readin' - there is much to learn.

> themselves unable to use them with the A640. The A640's 10mp
> sensor won't have the resolution of a good 10mp DSLR, but the
> resolution it does provide is probably about as good as the
> resolution you'd get from other 10mp P&S cameras using the
> same size sensor.

> performance. The reviews indicate that even though more
> pixels are squeezed into the same size sensor that is used by
> the 7mp sensor in the A620 (which I own), Canon has managed to
> do so without degrading the image with more noise.

On Feb 8, 11:48 am, "Paul D. Sullivan" <dudeb...@hotmail.com> wrote:
> Thanks for the clarification.
>
> I guess it is hard to find some general rule of thumb about which
> sensor size can properly handle what level of megapixels.
>
> I guess I was hooping for something like this, for example:
>
> "An Oly 4/3rds sensor is large enough to properly resolve a full
> 12 megapixels of image data - anything above that and you might
> start to lose detail because the sensor cannot properly
> accommodate more than 12 megapixels. In order to get higher than
> 12 megapixels and have those extra pixels be meaningful, they
> would have to increase the size of the sensor"

Well, the reason you don't get such an answer is probably that you
haven't defined the question carefully enough. There's not much
problem with these small sensors with lots of pixels, except for a)
noise, b) high pixel density which means the lens needs to resolve
many more line pairs/mm (at the sensor) than for eg a DSLR (which has
much larger pixels). The second problem is alleviated by the fact that
the sensor is smaller and thus the size of the image that must be
projected by the lens is also smaller, so that the lens doesn't cost
as much as one with equivalent resolution with a DSLR (or medium
format back!). On the other hand, I would have thought that not all
that much money and effort would be expended on making the lenses on
these compacts as sharp etc as possible, so probably some degradation
is also coming from there.

So, one cannot say "a sensor of size m cannot properly accomodate more
than n pixels", for the simple reason that you have not defined
"properly". If you say eg "this amount of noise is acceptable", then
probably a rough estimate can be found, or at least an upper limit for
n. But how do you plan to quantify it? Better to look at images and
decide.

There's not much else wrong with small sensors beside this noise
business (which can be a huge problem of course!), and of course there
are several advantages (cheaper and smaller lenses and cameras,
possiblity of having more versatile zooms with reasonable quality
etc). But a 10mp camera with a small sensor (such as the various
recent compacts) will always have more noise than a 10mp with a bigger
sensor (at least if they are both newer than, say, 2004).

>
> But I guess it isn't that simple.
>
> I do appreciate the reply. The more I read, the less that
> article seems to be on target. They generalized too much, I
> think.
>
> Canon certainly seems to feel that the sensor in their 10mp A640
> is fully capable of handling all 10mp of data properly, and
> Olympus seems to feel very confident that the size of its 4/3rd
> sensor is able to handle a full 8mp without any compromise in
> quality.

Well, as for handling the data, it's not a problem; the problem is to
actually collect it. But clearly if Canon did feel that the sensor
could handle 10mp with no problem, why do they produce cameras with
bigger ones? Do you really think that people pay 6-7000 euro for a
1Dmkwhateveritis just to be able to use their old lenses, if they
could get similar results otherwise? The differences, when they are
visible, can be huge, and so is the price difference!

>
> Anyway, I'll keep on readin' - there is much to learn.
>

On Feb 7, 8:46 am, "Paul D. Sullivan" <dudeb...@hotmail.com> wrote:
> I read an article on Cnet about how 8 or 10 megapixel cameras are
> not as good as they could be because their megapixel capability
> does not match sensor sizes well. If you buy a camera with 8 or
> 10 megapixels and they have some really small sensor, the extra
> megapixels will not be realized in terms of quality.
>
> Got me thinking - what are some of the best matches of megapixel
> capability to sensor size?
>
> How does the Canon a640 mentioned in the article hold up?
>
> How does the Canon S3 IS hold up?
>
> Olympus Evolt series with those 4/3rds sized sensors?
>
> Do Fuji, Kodak, Nikon and others have a good ratio on average, or
> are certain models better than others?
>
> I'd like to think that consumers aren't getting screwed when they
> want to get to an 8 or 10 megapixel camera, and wanted to ask
> some of ya'all here about it so I could learn what up.
>
> Any helpful answers would be appreciated.
>
> Thanks



It is the same situation as with film- "bigger is better". There used
to be an old saying- "nothing beats format size". While sometimes
people made higher acuity film for smaller formats, frequently the
film would gravitate up to larger formats.

With digicams bigger format size means larger aperture diameter at
same f/#, and collecting more photons during the same exposure time
period.

More expensive, of course. Silicon chips are more expensive when
larger, larger lenses needed for same field of view. But- better!

On Thu, 08 Feb 2007 06:56:35 GMT, "Paul D. Sullivan"
<> wrote:

>But I guess I'd further ask, what megapixel size is that sensor
>able to resolve properly? Is 8 mp a good fit for that 4/3rds
>sensor? Would it max out at 8 mp or would it also work well for
>10 mp or 12 mp too?

I think you're asking for objective answers to subjective questions.
How do you define "resolve properly"?
What is a "good fit"?
What does "work well" mean?

The answers will differ depending on who's asked. The end result is
that only you can answer these questions for yourself.
There are sites (such as, for example, dpreview.com) that offer sample
images from reviewed cameras, so that you can see for yourself how
these various cameras handle different conditions.
At such sites, you can see how well each camera handles the
compromises between different design goals; is 10MP on this camera
worth the trade-off in more noise? Or should you stick with the 8MP
model?
These are questions only you can answer; that's why there are so many
different makes and models out there; they offer different
combinations of compromises to attempt to meet the demands of as many
users as possible. But they aqre all compromises; there is no perfect
answer to your questions.

--
Massachusetts' former governor
Mitt Romney said Tuesday he will
announce his candidacy for the
GOP nomination for president next
week. He's a Mormon. It's not
expected to hurt him as long as
Rudy Giuliani is the candidate
with three wives.

Paul D. Sullivan wrote:
> That's a ton of data that I can't begin to understand. I wonder
> if you would be so kind as to help summarize it in simpler terms?
>
> For example, you indicate that the 4/3rds in the Oly Evolt is mid
> range. That makes sense and I can understand that.
>
> But I guess I'd further ask, what megapixel size is that sensor
> able to resolve properly? Is 8 mp a good fit for that 4/3rds
> sensor? Would it max out at 8 mp or would it also work well for
> 10 mp or 12 mp too?

For each camera, first compute the pixel pitch. The pixel
pitch gives the best indicator of the active area of each pixel.
To compute pixel pitch, look up the size of the sensor
in mm. Many cameras have only an obscure size related to
vidicon vacuum tube size from over 50 years ago.
Sensor size conversion is here:
http://www.clarkvision.com/imagedeta...Sensor%20Sizes

Let's say your camera has a 1/2.5 inch sensor. The above table
shows the real sensor size is 5.76 x 4.29 mm Now look up
the maximum size of images (not interpolated digital zoom)
the camera makes. For example the Canon S3 with a 1/2.5-inch sensor
gets 2816 x 2112 pixels. Take the size in mm times 1000 divide by
the number of pixels:

pixel pitch = long dimension in mm * 1000 / #pixels in long dimension

Canon S3 pixel pitch = 5.76 * 1000 / 2816 = 2.0 microns.

(there are 1000 microns per mm; the reason for the 1000 factor).

Now go to
http://www.clarkvision.com/imagedeta...ary/index.html

and plot a point in the gray band on Figure 2, 4 and 6 on the above
web page to get an idea of the camera performance, and how it
compares to other cameras. For example, on Figure 6, unity
gain ISO, 2-microns plots around ISO 100 in the gray band, the
bottom of performance. The unity gain ISO is the ISO where
1 digital number corresponds to 1 electron, so using higher ISO
is no help. This tells the high ISO performance of the camera.
If, for example, you want to take pictures indoors without flash of moving
subjects (baby's first step, kids and pets at play), you want unity gain
ISO to be as high as possible, with ISO 400 being a point where
images are reasonable quality (my opinion). In effect, ISO's higher
than the unity gain ISO is like "digital ISO." "Digital ISO,"
like digital zoom can be done in post processing and gains no
additional real image information. Camera manufacturers should
be able to publish unity gain ISOs on each camera in my
opinion.

Plotting points on Figure 2 will show image quality of a typical
well-exposed scene. You want to be in the 40 to 1 range or higher.
If the camera has ISO 50, the boost the numbers by 1.4x, so
ISO 100 giving signal-to-noise ratio (S/N) = 30 would give
30*1.4 = 42 at ISO 50 and good quality. If you do a lot of
editing (stretching, dodging and burning), you need higher
S/N. Also, if you want good shadow detail, you need good
S/N.

For total image quality, good lenses with high megapixel count
using larger pixels are best, but that makes for a larger (bulk and
weight) camera and higher cost. Only you can decide what
trade point you want/can afford.

Roger
>
> Thanks much.
>
>> While one can't prove there is an optimum, because the
>> trade space of size of a pixel versus performance just keeps
>> getting better with larger pixels, I personally think the
>> about 6 to 8 micron pixel pitch is the ideal size. That pitch
>> allows many pixels to be crammed into an small sensor
>> (e.g. about 8 megapixels in APS-C), while still collecting
>> enough photons for great performance, including high ISO
>> performance. Collecting about 50,000 photons per pixel
>> produces beautiful images, and the 6 to 6 micron pixel pitches
>> do that. This is the performance point of many DSLRs.
>
>

There is an easy way to calculate needed pixel pitch for
a "quality" camera.

That is to assume a "usual" f/number for the lens.

Assume that it will be near diffraction limited at
two stops or a little more from wide open. Calculate the
pixel pitch needed to resolve this without aliasing.
From that and the sensor size you get the needed
number of pixels.

Let's see .. assume an f/5.6 lens, and diffraction
limited at f/8. Assume a sensor size of 200x250 mm
..... and calculate. Ummmmm ... yummy.

Doug McDonald

Doug McDonald wrote:
>
> There is an easy way to calculate needed pixel pitch for
> a "quality" camera.
>
> That is to assume a "usual" f/number for the lens.
>
> Assume that it will be near diffraction limited at
> two stops or a little more from wide open. Calculate the
> pixel pitch needed to resolve this without aliasing.
> From that and the sensor size you get the needed
> number of pixels.
>
> Let's see .. assume an f/5.6 lens, and diffraction
> limited at f/8. Assume a sensor size of 200x250 mm
> .... and calculate. Ummmmm ... yummy.

You mean like this?
http://www.clarkvision.com/photoinfo/large_mosaics

Roger